journal of child language €¦ · downloaded: 19 feb 2016 ip address: 194.94.96.194 theakston et...

Journal of Child Languagehttp://journals.cambridge.org/JCL

Additional services for Journal of Child Language:

Email alerts: Click hereSubscriptions: Click hereCommercial reprints: Click hereTerms of use : Click here

Going, going, gone: the acquisition of the verb ‘go’

ANNA L. THEAKSTON, ELENA V. M. LIEVEN, JULIAN M. PINE and CAROLINE F. ROWLAND

Journal of Child Language / Volume 29 / Issue 04 / July 2002, pp 783 - 811DOI: 10.1017/S030500090200538X, Published online: 21 November 2002

Link to this article: http://journals.cambridge.org/abstract_S030500090200538X

How to cite this article:ANNA L. THEAKSTON, ELENA V. M. LIEVEN, JULIAN M. PINE and CAROLINE F. ROWLAND (2002). Going, going, gone: theacquisition of the verb ‘go’. Journal of Child Language, 29, pp 783-811 doi:10.1017/S030500090200538X

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/JCL, IP address: 194.94.96.194 on 19 Feb 2016

http://journals.cambridge.org Downloaded: 19 Feb 2016 IP address: 194.94.96.194

J. Child Lang. (), –. # Cambridge University PressDOI: .}SX Printed in the United Kingdom

Going, going, gone: the acquisition of the verb ‘go’*

ANNA L. THEAKSTON

University of Manchester

ELENA V. M. LIEVEN

Max Planck Institute for Evolutionary Anthropology, Leipzig

JULIAN M. PINE

University of Nottingham

CAROLINE F. ROWLAND

University of Liverpool

(Received July . Revised February )

This study investigated different accounts of early argument structure

acquisition and verb paradigm building through the detailed exam-

ination of the acquisition of the verb Go. Data from children followed

longitudinally between the ages of ; and ; were examined.

Children’s uses of the different forms of Go were compared with respect

to syntactic structure and the semantics encoded. The data are com-

patible with the suggestion that the children were not operating with a

single verb representation that differentiated between different forms of

Go but rather that their knowledge of the relationship between the

different forms of Go varied depending on the structure produced and

the meaning encoded. However, a good predictor of the children’s use

of different forms of Go in particular structures and to express particular

meanings was the frequency of use of those structures and meanings

with particular forms of Go in the input. The implications of these

findings for theories of syntactic category formation and abstract rule-

based descriptions of grammar are discussed.

[*] We would like to thank all the families who took part in the research reported here. Also,

thanks to two anonymous reviewers for their helpful comments on an earlier draft of this

paper. This research was funded by the Economic and Social Research Council, Grant

Numbers R and R. Address for correspondence: Anna L.Theakston, Department of Psychology, University of Manchester, Oxford Rd.,

Manchester, M PL, UK. tel : () -. fax: () -. e-mail :theaksto!psy.man.ac.uk

http://journals.cambridge.org


ET AL.

To understand how children attain adultlike verb use, two central questions

must be answered. First, how do they form verb representations that allow

them to use a verb in different sentence structures? Second, how do they

acquire the grammatical systems for marking tense and agreement in a

language necessary for the correct use of inflected verb forms? To explore

these questions, we must first consider how this information may be

represented in the adult grammar.

Verb semantics and syntactic structure representations in the adult grammar

The question of how verb meaning is represented in the adult lexicon is the

subject of much debate (e.g. Levin & Hovav, ), in particular how

variations in verb meaning, e.g. ‘I sneezed a napkin off the table’ vs. ‘I

sneezed ’ ; Goldberg, , are represented. Croft () outlines four models

of the adult lexicon concerned with this question: () the independent entries

model, where the two meanings are stored as independent lexical items with

no relation between them, () the polysemy model, where the two meanings

are stored independently, but speakers recognise a semantic link between

them, () the derivational model, where the prototypical meaning is stored in

the lexicon, but the second meaning is derived via the application of a lexical

rule, and () the pragmatic model, where the central meaning is stored in the

lexicon, but additional meanings are computed from general cognitive

principles relevant to the specific context of use.

Generative approaches typically assume that either a single lexical entry

is stored for each verb and alternative meanings are derived by applying

general rules (e.g. Levin & Hovav, ), or multiple but unrelated

entries are stored to represent different meanings (e.g. Pinker, ). In

many models alternative meanings have no status in the lexicon but rather are

computed ‘on the fly’ (Croft, : ). In contrast, constructivist models

assume polysemy in representation (Langacker, , ; Croft, in press

b). Lexemes are stored multiple times in a distributional network to encode

different meanings, with links between entries made on the basis of

phonological, semantic, or distributional similarity. Two central differences

between these approaches are the constructivist assumptions that in the adult

grammar, incorporating both idiomatic and more productive utterances, ()

whole phrases and individual words are stored as units in the lexicon, and ()

speakers have grammatical knowledge stored at many different levels of

abstraction. Frequently used constructions may be stored as units that can be

retrieved as lexical wholes, as well as being exemplars of a more abstract

construction. Thus, although adults have access to abstract grammatical

representations, much of adult language use may be tied to frequently



‘ ’

produced words and phrases stored as units in the speaker’s lexicon. For

example, adults are thought to have an abstract representation of the

concepts Subject, Auxiliary, and Verb, but differences in the phonological

reduction patterns in the production of the phrase I don’t know (e.g. I dunno)

suggest that whole phrases are often retrieved as single units (Bybee &

Scheibman, ). Which phrases are stored as units is determined by type

and token frequency of use.

This raises an important issue in language acquisition. If children first

acquire frequent phrases that are stored as units in the adult lexicon, it is

unclear whether they (a) have access to the underlying grammatical structure,

or (b) initially learn whole phrases and only later determine their grammatical

structure.

Verb representations and the use of inflection in the adult grammar

Generativists regard the use of inflection as a formal syntactic operation, and

assume that adult speakers operate with a single lexical entry for individual

verbs. Only irregular forms are stored in the lexicon occupying a cell in the

inflectional paradigm for that verb (Marcus, Pinker, Ullman, Hollander,

Rosen & Xu, ; Pinker, ). The full range of verb forms is computed

in the same way for every meaning of that verb, by either the application of

general inflectional rules, or the inheritance of irregular forms from an

existing verb meaning (Pinker, : ). Constructivists, in contrast, claim

that inflectional morphemes have semantic and phonological content, and are

stored in the lexicon. Different inflected verb forms therefore consist of

different combinations of meanings, and have different representations in the

lexicon (Langacker, ). Whether an inflected verb form is stored as a unit

or computed from an abstract schema is determined by the type and token

frequencies of particular combinations (Bybee, ), and might differ

according to the specific construction in which that form appears.

The acquisition of the verb lexicon

The question of how children acquire the range of meanings associated with

specific verbs and how this interacts with their acquisition of inflection is

complex. This paper is a longitudinal study of the acquisition of the verb Go

in the speech of English-speaking children between the ages of ; and

;." Go is acquired early (Tomasello, ), is very frequent in early speech

(Bloom, Miller & Hood, ; Clark, ; Naigles & Hoff-Ginsberg, ),

[] Terminology: throughout this paper, the form ‘Go’ (upper case ‘G’) will be used to referto the verb ‘Go’ as a whole while ‘go’ (lower case ‘g’) will be used to refer specifically to

the stem}non-inflected form of the verb; ‘structures’ will be used to refer to the syntacticstructures listed in Table , ‘meanings’ will be used to refer to the semantic categorieslisted in Table , and the phrase ‘different forms of Go’ will be used to refer to the formsgo, going}gonna, goes, gone, and went.



ET AL.

and can be used in different syntactic and to encode different

(see Tables and ). The of Go (go, going}gonna,goes, gone, went) are easily identified making it possible to examine children’s

understanding of the relation between them. The present study has two

aims; first to provide detailed documentation of the use of Go over a long

period of development, and second, to use the data to investigate whether

generativist and constructivist approaches to early verb learning can be

distinguished empirically. Before presenting the data, we will consider what

predictions can be derived from these approaches.

The acquisition of different meanings of Go. Although generativists assume

that children’s early speech reflects the workings of an abstract grammar,

they may argue that children will not have lexical rules governing argument

structure alternations until they have acquired a large number of verbs.

Initially, different verb meanings may be stored as separate lexemes from

which the verb’s argument structure is projected (Pinker, ), and may be

acquired separately. No specific predictions follow regarding their order of

acquisition. Constructivists assume children’s grammatical knowledge

initially consists of constructions based on high frequency forms in the input

(Tomasello, ; Pine, Lieven & Rowland, ). Only when they have

acquired a number of exemplars of a particular construction will they build

the more general schemas, such as the transitive construction (Akhtar &

Tomasello, ), that underlie adult language use. Children should first

acquire frequently modelled meanings of Go as lexically-specific con-

structions, and will only begin to build links between different meanings of

Go later in development when they have acquired a number of forms of Go

with a number of structures or meanings.

These two positions cannot be distinguished empirically as both predict

that initially, different of Go will exist independently, and nativists

make no predictions regarding order of acquisition. Can these approaches be

distinguished in terms of the early use of the different of Go?

The acquisition of the different forms of Go. The use of inflected verb forms

is often taken as evidence that children have an abstract knowledge of tense

and agreement. However, few researchers make specific predictions re-

garding the process by which children recognise that different lexemes are

exemplars of a single verb. This question becomes more complex when

considering verbs with multiple meanings.

Two generativist approaches to the acquisition of inflection will be

considered. The first assumes that children have acquired the knowledge

necessary to distinguish between different forms of Go before they enter the

two-word stage, and produce utterances like he go and he goes as optionally

inflected forms of Go (Wexler, , ). Thus, from the earliest stages,

children have the knowledge necessary to produce all forms of Go with any

structure or meaning used with another form of Go. The second approach



‘ ’

assumes a more gradual pattern of acquisition (Pinker, ; Radford, ).

Children may initially produce different forms of Go with different structures

or meanings without recognizing the relation between them. However, given

the assumption that a verb’s syntax is projected from its semantics, it seems

reasonable to assume that once children have produced two forms of Go in

the same structure, they will recognise a semantic relation that will enable

them to enter them in a single paradigm. The phonological similarity

between paradigms associated with individual meanings or structures should

also facilitate the construction of a unified verb paradigm. Together, these

processes will rapidly enable children to use any form of Go with any

structure or meaning used with other forms of Go.

From a constructivist perspective, children do not initially have a knowl-

edge of inflection, nor are they expected to rapidly acquire verb-general

inflectional schemas. Instead, the acquisition of different forms of Go with

different structures or meanings will reflect their frequency of use in the

input. As the relative frequency of forms of Go is likely to differ across

constructions, it may take time for children to develop the knowledge

necessary to produce all forms of Go with all structures and meanings used

with Go.

It should be possible to distinguish empirically between generativist and

constructivist approaches to the development of verb paradigms. From a

generativist perspective, only minimal overlap in the structures or meanings

found with different forms of Go is needed before children have the

knowledge necessary to build a verb paradigm. Although the input frequency

of different forms of Go might influence their initial acquisition, once

acquired, their frequency of use with different structures or meanings in the

input should not affect children’s ability to produce them all with any

structure or meaning acquired with Go. From a constructivist perspective,

the development of a general verb paradigm available for use across different

structures and meanings will be a much lengthier process because it depends

on the accumulating type and token frequencies of different forms in

different constructions. Therefore, children’s use of different forms of Go

will be limited, and the acquisition of individual forms with specific

structures and meanings will reflect the frequency of those forms in those

constructions in the input.

Issues of sampling. Although generativist approaches should predict that

children will produce the different forms of Go, once acquired, with all

structures and meanings used with Go, the actual likelihood that children will

produce these structures and meanings with different forms of Go will be

determined by a combination of their grammatical knowledge, the preferred

forms of expression in the language,# and the context in which speech is

[] Some structures are fully grammatical, but nevertheless rarely appear in adult speech.



ET AL.

sampled. While the absence of a specific structure or meaning with different

forms of Go in children’s speech might mean that they have not acquired that

structure or meaning with those forms of Go and have a lexically-specific

knowledge of syntax, it might instead reflect the preferred forms of ex-

pression. Furthermore, although constructivists predict that frequently

modelled forms in the input will be acquired earlier by children than low

frequency forms, this pattern of results could reflect sampling considerations.

Children with adultlike grammatical knowledge may produce high frequency

forms from the input earlier in sampled speech over the course of de-

velopment than lower frequency forms.

Given these considerations, we cannot distinguish empirically between the

different approaches with respect to the order of acquisition or the

generalization of use of different forms of Go across different structures or

meanings without both an exhaustive speech sample (to avoid sampling

problems) and experimental studies (to elicit the use of structures that are

rarely found in spontaneous speech). Further analyses are necessary. The

method adopted here is to compare verb use in specific syntactic structures

(e.g. VPP) in matched samples of mothers’ and children’s speech, takinginto account the acquisition of specific lexemes. Using matched samples

removes the possibility that children’s verb use appears more limited because

they speak less than adults, and ensures that if adults and children operate

with the same underlying grammatical knowledge, the likelihood that

particular verb uses will appear in the children’s and adults’ speech is equal.

Moreover, it is within specific syntactic structures that constructivists would

first expect generalization between forms of Go to occur (construction-

specific vs. more general paradigms).

To summarize, the aims of the present study are to examine to what extent

() uses of different forms of Go are syntactically or semantically distinct, ()

children generalize knowledge of structures and meanings across different

forms of Go, and () children’s use of different forms of Go differs from their

use in the input. The following predictions were derived:

Constructivist. () Children will learn high frequency constructions from

the input with individual forms of Go earlier in development than low

frequency constructions, () little generalization between forms of Go will

occur, reflecting the fact that they are initially associated with different

constructions, and () in specific constructions, the use of particular com-

plements in the children’s speech will be less diverse than in the input (e.g.

fewer different prepositions occurring in prepositional complements in the

children’s speech than in the input), reflecting the greater degree of

abstraction underlying adult speech.

Generativist. () From early in development, children will have the

knowledge needed to produce all forms of Go with all available structures and

meanings, but high frequency constructions found in the input will be



‘ ’

. Syntactic structures used with the verb ‘G ’a

. V (Go)adverb I go here; He went fast. V (Go)PP It goes in there. ‘Go’infinitive}gerund He’s going to hit it ; I’m going shopping. ‘Go’ within infinitival complement I want to go home. Other two verb utterances Let me go out; I’ll go and put it on. Yes–no question

(with subject auxiliary inversion)

Are you going?

. Tag questionb He’s going to the shops, isn’t he?. Wh-question Where’s it gone?. ‘Go’ with no complement It’s gone; He’s going

. Reversed word order(all cases where word order is reversed)

There you go; Off we go

a In declaratives, auxiliaries and auxiliary-like verbs followed by an -ing participle (e.g. stop

going, keep going) were ignored for the purpose of utterance categorization. These utterances

were coded as single verb utterances rather than two verb utterances.b Tag questions were coded separately because they create a third type of question, and

involve coordinating knowledge of the person marking in the main clause (which can be

encoded on the main verb – it goes there, doesn’t it, or on an auxiliary it does go there, doesn’t

it, and thus interacts with knowledge of marking on the main verb) with the appropriate

auxiliary form for use in the tag. However, there were only a very small number of tag

questions produced by any of the children so the decision to code these utterances in a separate

category does not impact on the overall analysis.

sampled earlier than low frequency constructions, () there will be little

evidence of generalization between forms of Go if that reflects the preferred

forms of expression in the language, and () in specific constructions, the

diversity of use of particular complements will be similar in the children’s

speech and the input, reflecting the abstract nature of the grammar under-

lying both adult and children’s speech.

Note that it is difficult to empirically differentiate the two approaches with

respect to predictions () and (), while prediction () ought to be testable.

The difficulty in empirically distinguishing between these two seemingly

diametrically opposed approaches will be taken up again in the Discussion.

Subjects

The participants were eleven of twelve children who took part in a

longitudinal study of early language development. One child did not produce

all forms of Go and was excluded. The children were from predominantly

middle class, monolingual English-speaking families and were recruited

through newspaper advertisements and local nurseries. They were first-

borns and were cared for primarily by their mothers. At the beginning of the



ET AL.

study the children ranged in age from ;. to ;. with MLUs ranging

between . to . in morphemes. The Manchester corpus (Theakston,

Lieven, Pine & Rowland, ) is available on the CHILDES database

(MacWhinney, ).

Procedure

The children were audiotaped in their homes for an hour on two separate

occasions in every three week period for one year. They engaged in normal

everyday interaction with their mothers. The data were orthographically

transcribed using the CHILDES system (MacWhinney & Snow, ;

MacWhinney, ).

The children’s speech corpora

Using the CLAN programmes (MacWhinney & Snow, ; MacWhinney,

) the transcripts were searched for all instances of the different forms of

Go (go, going, gonna, goes, gone, went). Self-repetitions, imitations, incomplete

or partially intelligible utterances and routines (e.g. nursery rhymes,

counting) were excluded from the children’s corpora. The data for each child

were divided into eight time periods of approximately six weeks, seven

comprising four one-hour recordings and the last time period comprising six

one-hour recordings.

Syntactic development. For each form of Go, the utterances were coded

according to the structures listed in Table . One tenth of the children’s

utterances taken equally across the children, half taken from the beginning

and half from the end of the study, were coded by a second investigator.

Inter-coder reliability measured using Cohen’s kappa resulted in a coefficient

of ±. For each form of Go, children were assumed to have acquired aparticular structure once they had produced two utterances with that form of

Go in that structure (e.g. go awaygo in¯ goAdv ; goes in that doorgoesin there¯ goesPP).

Semantic development. For each form of Go, the utterances were coded

according to the meanings listed in Table . Where utterances were

ambiguous, linguistic context (i.e. preceding and following utterances) was

used to determine the child’s meaning. One tenth of the children’s utterances

taken in equal numbers from each child, half from the beginning and half

from the end of the study, were coded by a second investigator. Inter-coder

reliability measured using Cohen’s kappa resulted in a coefficient of ±. Foreach form of Go, children were assumed to have acquired a given meaning

once they had produced two utterances with that form of Go with that

meaning (e.g. it’s goinggoing to hospital¯ going}movement ; that wentbangit went bang¯went}sound).

Lexical diversity. To establish the extent to which children generalized use



‘ ’

. The semantic senses encoded by the verb ‘G ’a

. Movement(movement from one location to another)

He’s gone to school; I’m going in the car

. Belong}fit(person or entity belongs in a specific

location, fits in a specific location)

Does that piece go there?; The book goes

on the shelf

. Disappearance(person or entity is no longer present}missing without implication of

prior movement)

Where’s that book gone? (when looking

for it)

. Future intentb

(infinitival)

I’m going to dance; She goes to get him

from school

. Ambiguous(no context)

. Encourage Go on, you can do it. Sound

(sound, noise or speech)

It went bang; Dogs go woof

. Happening What’s going on?; What went on here?. Give up possession Let go

. Specified motion(at specific location)

Go like this with your foot ; The tape’s

going round and round

. State}outcome It’s gone wrong; It’s going cold. Order of occurrence You go first. Aim for He’s going for the red one; I went for the

black.

. Searching}reading etc. Let’s go through this book; I wentthrough that pile

. Increase}decrease(size, value)

The price went up; That balloon’s going

down

. Resume Go back to the jigsaw; You’re going backto the toys

. Catch illness You’ll go down with a cold; He wentdown with flu

. Dislike Have you gone off chips?; I went offjigsaws

. Prepare Get the dinner going

a All utterances with reversed word order (e.g. there you go) were excluded from the semantic

analyses because they were frequently ambiguous in their meaning. In all cases, surrounding

utterances were examined to provide contextual information to inform the semantic classi-

fication. (e.g. The balloon’s going down would only be coded as Increase}decrease rather thanMotion, if it was clear that the utterance referred to the deflation of the balloon, and not the

movement of the balloon, for example, down the stairs.)b There is an issue with respect to the semantics encoded by infinitival structures. While the

progressive verb form ‘going’ has, over time, grammaticised and can be used to express

intent or future events without necessarily encoding any form of movement (Bybee et al.,

) (e.g. I’m going to think about it), other verb forms have retained the reference tomovement in this structure (He’s gone to think about it ; I went to think about it). Some

uses of going can also be ambiguous in this respect and may refer to either movement or

a future event (e.g. I’m going to open the door). The lack of detailed contextual information

meant it was impossible to differentiate these meanings, therefore for the purposes of this

study, all infinitival utterances are classified as examples of the semantic category ‘Future

intent’.



ET AL.

of the individual lexical items associated with specific structures (e.g. the PPs

used in VPP utterances) across different forms of Go used in thosestructures, and to compare the children’s use with use in the input, the range

of lexical items produced in particular structures with each form of Go was

documented. The structures examined were PP complements, infinitival

complements and Wh-questions.

The mothers’ speech corpora

The mothers’ data were searched for all exemplars of Go and the children’s

data were compared with the input for the use of different structures and

meanings, and for the diversity of use of lexical forms within specific

structures.

Syntactic structures. It was not possible to directly compare the relative

proportional use of different structures in the children’s speech with their use

in the input because the mothers produce a much greater proportion of

complex structures (e.g. two clause utterances, questions) than the children,

even though it is reasonable to assume that children learn to produce simple

structures (e.g. VPP) from many possible sources in the input includingcomplex sentences. Therefore, for purposes of comparison with the

children’s speech, the mothers’ data from the first eight tapes were coded for

a subset of the structures listed in Table (VAdv, VPP, Go as infinitive,Yes}No question, Wh-question, no complement, and word order reversals).Individual phrases with Go, rather than whole utterances, were coded.

Questions were coded twice, once as a Yes}No or Wh-question, and onceaccording to the complement of the verb Go. The proportional use of each

structure was then calculated.

Verb semantics. For purposes of comparison with the children’s speech,

the mothers’ data from the first eight tapes were coded according to the

meanings listed in Table . The proportional use of each meaning was then

calculated.

Lexical diversity. Two sets of corpora were selected for the purposes of

comparison with the children’s speech. First, to compare lexical diversity in

the children’s speech and in the input, each mother was matched with her

child for the number of utterances produced with each form of Go for the

whole year. The mothers’ utterances were taken consecutively beginning

from the first transcript until the correct number of utterances were selected.

On four occasions, the total number of maternal utterances available for a

particular form of Go was less than the number of utterances produced by her

child for that form. Second, to investigate whether frequency in the input

was related to the order of acquisition of particular lexical forms in the

children’s speech throughout the year, the input data from the first eight

tapes were examined.



‘ ’

Correlation analyses

All correlational analyses are based on the mean use of particular structures

or meanings with different forms of Go in the children’s speech and in the

input. In each case, pairwise correlations were conducted, reflecting all

possible comparisons between the mothers or children, to establish

whether similarities existed between mothers or children. These correlations

show that the use of particular structures or meanings is similar across

mothers and across children, thus justifying the decision to report mean

values.$ A summary of the correlations conducted is given with each

analysis detailing the number of items included in each correlation, and the

proportion of correlations where the correlation coefficient was above a given

value.

The proportional use of the different forms of Go

Table shows that there are differences in the children’s proportional use of

each form of Go. The form go accounts for the greatest proportion of the data,

going and gone are relatively frequent, but goes and went account for a small

proportion of the data.%

This is unsurprising as goes and went are probably relatively infrequent in

adult language while go is the most common form to be used as the verb

complement of other verbs (e.g. I want to go home) and thus accounts for a

greater proportion of the data.& The question of interest is whether the forms

of Go differ with respect to the structures and meanings with which they

appear as well as in their frequency of use.

Syntactic development

Differential use of syntactic structure. Only those structures produced by at

least of the children were included in the following analyses. From Table

, these were () Vadverb, () VPP, () infinitives and gerunds, () two-

[] For each comparison between the mothers and children’s speech where pairwisecorrelations are reported, individual dyad analyses were also carried out to check whether

relationships observed in the combined data were also found between individual mothers

and their children. For clarity of presentation, and due to space limitations, the mean

values are reported, but in all cases, a minimum of of the individual dyad correlationswere significant.

[] There was only a single instance of overregularized goed in the entire corpus (Anne, tape, a& goed up). As it is difficult to determine the meaning of this utterance (even takinginto account the linguistic context), and Anne did not produce any further exemplars of

goed, nor any exemplars of the irregular past tense form went until much later in the study

(tape ), it is unclear how this apparent use of goed relates to other forms of Go. Thisutterance was not analysed further.

[] Although other forms of Go can also be used as the complement of other verbs (e.g. I likegoing home), utterances of this type were exceedingly rare in the data.



ET AL.

. The proportional use of each form of ‘G ’

Number of

utterances

Proportional use of each form of ‘Go ’

go going goes gone went

Anne ± ± ± ± ±Aran ± ± ± ± ±Becky ± ± ± ± ±Carl ± ± ± ± ±Dominic ± ± ± ± ±Gail ± ± ± ± ±Joel ± ± ± ± ±John ± ± ± ± ±Liz ± ± ± ± ±Nicole ± ± ± ± ±Warren ± ± ± ± ±

Mean ± ± ± ± ± ±

. The mean proportional use of syntactic structures produced by +

children with each form of ‘G ’

Adv. PP

Infinitive}gerund

Two

verb

Yes–no

question

Wh-

question

No

complement

go ± ± ± ± ± ± ±going ± ± ± ± ± ± ±goes ± ± ± ± ± ± ±gone ± ± ± ± ± ± ±went ± ± ± ± ± ±

verb utterances (excluding infinitivals), () Yes–no questions, () Wh-

questions and () no complement utterances. Word order reversals () were

excluded because although in principle, this structure is available for use

with all forms of Go, ±% of the children’s utterances were with go or goes(there was only one exception), and therefore they do not provide a useful

indicator of children’s productive knowledge of syntax. To allow comparison

across forms of Go with respect to the relative use of each structure, Go

within infinitival complement utterances () were also excluded from this

analysis. This structure is highly frequent with go ( examples overall) but

is produced only once by each of four children with any other form of Go, all

of which are grammatically incorrect (e.g. I need going). Although other

structures are also ungrammatical with some forms of Go (e.g. Wh-questions

of the form Where’s X goes}went), all other structures were produced with atleast four forms of Go and were included in the analyses.

Table shows the mean proportional use of each structure with each form

of Go (N¯ : forms of Go¬ structures; % of the correlations were



‘ ’

above ±). The data suggest that there are differences between the forms ofGo with respect to their mean proportional frequency of use in these

structures. The most frequent forms used with infinitival complements are go

and going, goes is the most frequent form in adverb structures, gone is the

most frequent form in Wh-questions and no complement structures but the

least frequent form in adverb and PP structures while went is the most

frequent form in PP structures. Thus, the least frequent forms, goes, gone and

went, are not just less frequent than go and going but are also used in different

structures.

Given these differences between forms of Go, it is unclear whether the

children have adultlike grammatical knowledge and can use the different

forms in a range of structures, or whether the children’s knowledge is of a

more limited nature and related only to individual forms of Go. The mean

proportional use of structures with different forms of Go in the children’s

speech was compared with their mean proportional use in the input (N¯ : forms of Go¬ structures; % "±). The correlation was significant(r¯±, df¯, p!±) showing that there is a strong relationshipbetween use of individual forms of Go in particular structures in the input

and in the children’s speech.

Order of acquisition of syntactic structure. Overall, the children’s pro-

portional use of structures with different forms of Go closely resembles adult

use, but it is not clear if this provides an accurate representation of their

knowledge across the whole year. The mothers use all forms of Go in all

structures from the beginning of the study (tapes –), but the children may

take time to acquire the structures used with each form of Go. To examine

this possibility, structures were ranked according to the time period when

they were acquired for each child with each form of Go. Table shows the

mean rank order of acquisition of structures with each form of Go (N¯ : forms of Go¬ structures; % "±) and the approximate time periodin weeks between the start of the study and the acquisition of each form in

each structure.

There is a considerable delay between the first use of a structure with one

form of Go and its use with the full range of available verb forms. For

example, go is the first form used with adverb complements, going is used in

this structure weeks later, while goes and gone are produced approximately

weeks later than go. In contrast, the less frequent form gone is the first

produced in Wh-questions and no-complement structures, but there are then

considerable and differing time periods before other forms of Go are used in

these structures.

The data show that the differences in the proportional use of each structure

across different forms of Go are reflected in the time at which each form is

acquired in each structure. The children show little evidence that they are

able to produce all the forms of Go they have acquired in all the structures



ET AL.

. The mean rank order of acquisition of syntactic structures producedby + childrena

Wkb Adv. PP

Inf. &

gerund

Two

verb

Yes–no

qu. Wh-qu.

No

complement

– gonego

go

– gogone

go

– goinggo

going

go

going

– goinggone

goes

– goinggone

goes

– wentgone

going

went

– gonego

going

goes

goes

– wentgoes

went

went

a No other syntactic structures were acquired by the children during the study.b Approximate number of weeks into study when form acquired.

available to them. For example, by – weeks into the study, gone, and go

are used with no complement, yet gone is not used in many of the other

structures used with go until later in development. By the end of the study,

there is evidence that the children can use most forms of Go in most

structures, but this suggests a gradual pattern of acquisition.

Input frequency and the order of acquisition of syntactic structure. To address

the question of why children learn to use different forms of Go in some

structures sooner than others, the input data were examined to investigate

whether the frequencies of use of each form of Go in each structure predicted

their order of acquisition in the children’s speech. The mean frequencies of

use of each form in each structure in the input (N¯ : forms of Go¬



‘ ’

structures; % !±) were compared with their mean rank order ofacquisition in the children’s speech. The rank order correlation between

mean frequency of use in the input and mean order of acquisition in the

children’s speech was significant (ρ¯®±, df¯, p"±), and showsthat the order of acquisition of different forms of Go in different structures

in the children’s speech reflects their frequency of use in the input.

Semantic development

The children’s use of the verb Go was then examined with respect to the

meanings encoded by each form to establish whether differences in the

structures used with each form reflect differences in the semantics encoded.

For each form of Go the utterances were categorized according to the

meanings listed in Table .

Differential use of verb semantics. Only those meanings used by at least

of the children were examined. These were movement (), belonging (),

disappearance (), intent (), encourage () and sound (). However,

‘encourage’ is used only with go so this category was excluded from the

following comparisons. Table shows the mean proportional use of each

. The mean proportional use of each form of ‘Go ’ in differentsemantic structures produced by + children

movement belong disappear intent sound encourage

go ± ± ± ± ± ±going ± ± ± ± ± goes ± ± ± ± ± gone ± ± ± ± went ± ± ± ± ±

form of Go with each meaning (N¯ : forms of Go¬ structures; %"±). The reported means suggest there are differences between forms ofGo in terms of the meanings they are used to encode. The forms go, going, and

went are commonly used to encode movement, goes is mainly used to encode

belonging, and gone to encode disappearance. Going and, to a lesser extent,

go are used to encode intent more frequently than the other forms.

The differences observed in the children’s use of different meanings with

the different forms of Go suggest that they may not be operating with a single

lexical entry for the verb Go but rather with a number of separate or partially

related entries representing each form. To investigate this further, the mean

proportional use of meanings with different forms of Go in the input (N¯ : forms of Go¬ structures; % "±) was compared with their meanproportional use in the children’s speech. The correlation between the mean

proportional use of meanings in the input and the children’s speech was



ET AL.

significant (r¯±, df¯, p!±) showing that there is a strongrelationship between the use of different forms of Go to express particular

meanings in the input and in the children’s speech.

Order of acquisition of verb semantics. Overall, the children’s proportional

use of meanings with different forms of Go closely resembles adult use, but

it is not clear if this provides an accurate representation of their knowledge

over the whole year. The mothers use all forms of Go with all meanings from

the start of the study (except goes and went to encode disappearance), but the

children may take time to acquire the meanings used with each form of Go.

To examine this possibility, meanings were ranked according to the time

period when they were acquired for each child. Table shows the mean rank

. The mean rank order of acquisition of semantic structures in thechildren’s speecha

Weeks into

study movement belong disappear intent sound

– gonego

going

– go– go

goes

gone

going

– go– went– gone– going

went

goes

go

– goinggone

went

goes (error)

goes

a No other semantic structures were acquired by the children during the study.

order of acquisition of meanings with each form of Go (N¯ : forms ofGo¬ structures; % "±) and the approximate time period betweenthe start of the study and the acquisition of each form with each meaning.

There is a considerable delay between the first use of a meaning with one

form of Go and its use with other forms. For example, go and going are used

to encode movement from the beginning of the study but the children do not

use this meaning with gone and goes until – and – weeks into the

study, although they learn these forms much earlier. Even went, which is



‘ ’

learned later than goes, is used to encode movement earlier. Similar patterns

of development can be observed with respect to belonging, disappearance,

intent and sound which suggests that the children’s understanding of verb

meaning may initially be restricted to individual forms of Go and only slowly

developing towards adultlike knowledge.

Input frequency and the order of acquisition of verb semantics. To address the

question of why children learn to use different forms of Go with some

meanings sooner than others, the input data were examined to investigate

whether the frequencies of use of each form of Go with each meaning in the

input predicted their order of acquisition in the children’s speech. The mean

frequencies of use of each form with each meaning in the input (N¯ : forms of Go¬ structures; % "±) were compared with their rankorder of acquisition in the children’s data. The rank order correlation

between mean order of acquisition in the children’s speech and mean

frequency in the input was significant (ρ¯®±, df¯, p!±), andshows that the order of acquisition of different forms of Go with different

meanings in the children’s speech reflects their frequency of use in the input.

Lexical specificity in the use of different syntactic structures with each verb

form. Although the documented pattern of use of structures and meanings is

consistent with a gradual pattern of acquisition, the apparently late ac-

quisition of some structures or meanings with particular forms of Go may

reflect sampling limitations or the preferred forms of expression in the

language. Therefore, we examined the extent to which PP, infinitival, and

Wh-question structures show lexical diversity in the speech of the mothers

and the children using matched samples of speech. If children’s knowledge

of the relation between forms of Go is adultlike, the diversity of use of lexical

forms in these structures with the different forms of Go will be similar

between mothers and children. If the children have a less unified rep-

resentation of Go, their use of these structures is likely to be less diverse than

use in the input.

PP complements

Lexical diversity of PPs in matched samples. The number of different

prepositions used with each form of Go was calculated for each mother and

child. T-tests revealed that there were no differences between the mothers

and the children in the number of different prepositions used with each form

of Go with the exception of gone, where the mothers used a significantly wider

range of PP complements than the children (t¯®±, df¯, p!±).However, the children produced a wider range of PP complements with go

than with gone (go M¯± ; S.D.¯± ; gone M¯±, S.D.¯±), whichshows that the restricted use of PP complements with gone does not reflect a

lack of lexical knowledge. Although by the end of the year, the children could

use a similar range of different PP complements as their mothers with most



ET AL.

. The mean rank order of acquisition of PP complements producedby + children with different forms of ‘Go’a

Weeks into

study INX ONX TOX UPX LIKEX

– go– go

– going– go

goes

go

– goinggoing

gone

go

gone

– goesgoing

went

– wentgone

went

goes

went

– goesgoing

gone

goes

a No other PP complements were acquired by the children during the study.

forms of Go, they seemed unable to generalize their knowledge to the form

gone.

Order of acquisition of PPs. Those PP complements used by at least of

the children (in X, on X, to X, up X, like X ) were ranked according to their

order of acquisition with the different forms of Go for each child. Table

shows the mean rank order of acquisition of each PP in the children’s speech.

There were differences in the order and stage of acquisition of the PPs with

different forms of Go, which suggests that the children’s knowledge of PP

complements may be tied to individual forms of Go (e.g. onX is acquiredearlier with going than with goes, but likeX is acquired earlier with goes thanwith going). Although some prepositions were acquired later than others, this

does not explain why there were differences in the use of some prepositions,

once acquired, between forms of Go (e.g. go inX vs. gone inX).Input frequency and the order of acquisition of PPs. The input data were

examined to investigate whether the frequencies of use of each PP with each

form of Go in the input predicted their order of acquisition in the children’s

speech. The input data from the first eight tapes were searched for all



‘ ’

. The mean number of different infinitival verb complementsproduced with each form of ‘Go ’

Mean no. of verb

complements produced in

the children’s speech

Mean no. of verb

complements produced

in the input

go (incl. ‘ to’) ± (S.D.¯±) ± (S.D.¯±)going ± (S.D.¯±) ± (S.D.¯±)goes ± (S.D.¯±) ± (S.D.¯±)gone ± (S.D.¯±) ± (S.D.¯±)went ± (S.D.¯±) ± (S.D.¯±)

exemplars of the selected PPs and the frequency of each PP with each form

of Go was calculated. The mean frequencies of different PPs with different

forms of Go in the input (N¯ : forms of Go¬ PPs; % "±) werecompared with their mean rank order of acquisition in the children’s data (N

¯ : forms of Go¬ PPs; % "±). The rank order correlation wassignificant (ρ¯®±, df¯ ; p!±), showing that the frequency ofdifferent PPs with different forms of Go in the input is a good predictor of

their order of acquisition in the children’s speech.

Infinitival complements

For go, there were many instances in the children’s speech where it was

unclear whether the utterance should be classified as an infinitival comp-

lement of go, or as a coordinated structure where the conjunction and has

been omitted (e.g. I go get that is ungrammatical and could mean either I’ll

go to get that or I’ll go and get that). For purposes of comparison with the

input, the children’s data were searched for all complements of go produced

with the infinitival marker to as these are the only forms that can be clearly

interpreted as infinitives.

Lexical diversity of infinitival complements in matched samples. Table

shows the mean number of different infinitival complements produced with

each form of Go in the children’s speech and in the input. T-tests revealed

that the mothers produce a wider range of verbs with gone than the children

(t¯®±, df¯, p!±) whereas the children produce a wider range ofverbs with go than the mothers (t¯±, df¯, p!±). The similaritiesbetween the mothers and children in terms of the number of infinitival

complements produced with goes and went reflect the fact that very few

complements are produced by either group with these forms of Go.

Order of acquisition of infinitival structures. As the mothers produce very

few infinitival complements with go, it seems unlikely that the children have

learned these forms directly from their use in the input. Instead, they may

have learned a range of infinitival complements with going, and be



ET AL.

generalizing their use to go. The data were examined to determine the stage

at which the children acquired the frames gotoV and goingtoV.' Onaverage, the children who produced both frames acquired goingtoVearlier than gotoV (goingtoV –, and gotoV – weeks intothe study). The apparent generalization from going to go indicates that by this

relatively late point in development the children recognise a phonological,

semantic and}or distributional relationship between these forms. However,their restricted use of gone in infinitivals, in comparison with use in the input,

suggests that they are unable to generalize their knowledge of infinitival

complements from either going or go to gone. Thus, it appears that the

children do not yet have a complete understanding of the relation between

gone and other more frequent forms of Go.

Input frequency and the acquisition of infinitival complements. To investigate

whether the frequency of use of different infinitival complements with

different forms of Go in the input might influence their order of acquisition

in the children’s speech, input data from the first eight tapes were examined

and the frequency of use of each complement verb with each form of Go

calculated.

E For go, of the mothers who produced infinitivals, only one verb

(sleep) was produced by all mothers and accounted for a mean of

±% (range ±–%) of go infinitivals. For children sleep wasthe first or second verb to be produced with go and accounts for a mean

of ±% (range –%) of the children’s go infinitivals.E For going, the most frequent verb complements, calculated across

the mothers’ data, were ranked for order of acquisition with going in

the children’s speech. The rank order correlation between mean

frequency in the input and mean order of acquisition in the children

was significant (ρ¯®±, df¯, p!±), which suggests that thefrequency of different verb complements with going in the input

provides a good predictor of their order of acquisition in the children’s

speech.

E For goes, two different complement verbs were produced in the input

and only one child produced goes in this structure. Seventyfive percent

of the verb tokens in the input were accounted for by a single verb

(sleep), the only verb acquired by that child.

E For gone four different complement verbs appeared in the input. All

children produced this structure. One verb (sleep) accounted for the

majority of uses in the input (±%) and was the only verb acquiredby of the children (the remaining children produced just one other

verb each).

[] Acquisition was defined as the point at which a particular child produces two differentverbs with a particular form of Go.



‘ ’

E For went, four different complement verbs appeared in the input. Only

two of the children produced infinitivals with went (one and two verbs

only), but both acquired the verb see which accounts for % of

infinitivals produced in the input.

Thus, the data suggest that () although the children may have productive

use of the infinitival structure with going and go, their use of these forms

seems to be influenced, initially at least, by the frequency of particular

complement verbs in the input, () the children’s production of infinitival

structures with goes, gone and went is very limited (in the case of gone, their

use is more limited than in the input), and may reflect lexical learning tied to

high frequency verb complements used with each form in the input, and ()

there is little generalization between forms of Go (with the possible exception

of going and go at a relatively late stage in development).

Wh-questions

Wh-questions are interesting because the correct use of different Wh-words

with different forms of Go requires an understanding of the complex systems

governing tense and agreement. Sentences with Wh-subjects mark tense on

either the auxiliary, or the main verb, and therefore all forms of Go can

appear in simple questions of this type (e.g. What can go there?, What’s going

there?, What goes there?, What’s gone?, What went here?). In contrast, most

other Wh-questions mark tense on the auxiliary. Single clauses containing a

finite verb and an auxiliary (e.g. Where’s it goes?, Where’s it went) are

ungrammatical, therefore, the forms goes and went require the use of more

complex sentence frames to appear in questions of this type (e.g. Where’s it

go?,( Where’s it going?, Where’s it gone? vs. Where do you think it goes?, Where

do you think it went?). The issue to be addressed is to what extent children’s

early use of Wh-questions with different forms of Go illustrates an ability to

generalize knowledge across forms through an understanding of tense and

agreement.

Lexical diversity of Wh-questions in matched samples. The children’s data

and the matched samples from the input were searched for all Wh-questions

and the number of different Wh-words used with each form of Go calculated.

T-tests showed that the mothers produced a significantly greater number of

different Wh-words with go, going, and gone than the children (go, t¯®±,df¯, p!± ; going, t¯®±, df¯, p!±, gone, t¯®±, df¯, p!±). The similarities in the number of Wh-words produced withgoes and went reflect the fact that neither the mothers nor children produced

more than a mean of ± Wh-words with either form.

[] In the utterance Where’s it go?, the ‘ s ’ is assumed to be the cliticized remnant of does, i.e.Where does it go?.



ET AL.

. The rank order of acquisition of Wh-questions with each form of‘Go ’a

Weeks into

study where what who which how why whose when

– gone– go

– going– going– goes

go

– goinggoes

go

goes

gone

go

go

going

go

going

went

going

went

going

a No other Wh-forms were acquired by the children during the study.

These differences in Wh-question use could reflect the children’s lack of

knowledge of particular Wh-words. Therefore, the mean number of different

forms of Go produced with each Wh-word was calculated for the mothers

and children who produced that Wh-word (and therefore know the Wh-word

in question) to examine whether the mothers generalized use of each Wh-

word across forms of Go more than the children. The mothers produced a

wider range of forms of Go than the children with of the Wh-words

examined, which is consistent with the suggestion that even when children

have learned a particular Wh-word, they are less able to generalize their

knowledge of that Wh-question across forms of Go than the mothers.

Order of acquisition of Wh-questions. The children’s data were ranked in

terms of the order of acquisition of different Wh-words with different forms

of Go. Table shows the mean rank order of acquisition of Wh-words with

each form of Go (N¯ : forms of Go¬ Wh-words; % "±).Table shows that although the children acquired different Wh-words at

different stages in development, they did not produce given Wh-words with

different forms of Go at the same stage of development once those Wh-words

were acquired (e.g. where is acquired early with gone but not until later in

development with other forms of Go). Of particular interest is the slot-and-



‘ ’

frame appearance of the children’s where questions. Means of ±% (S.D.¯±), ±% (S.D.¯±) and ±% (S.D.¯±) of the children’s wherequestions with go, going and gone respectively consist of the two frames Where

X V? (i.e. Where (did}does etc.) X V?) and Where’s X V?. Although theseframes may have been acquired separately from the input, the children’s use

of Where questions with goes suggests that they may have made some

connection between forms of Go. Of the five children who produce where

questions with goes (M¯± utterances; range –), three only produce theframe Where X goes? while the other two only produce the frame Where’s X

goes?. The use of either frame is ungrammatical because both require a non-

finite form of Go. Although Where X goes? could be acquired from embedded

clauses (e.g. Is that where it goes?), those children who produce Where’s X

goes? must be using language productively as these forms are not present in

the input. However, where questions with goes are produced very late in the

study (– weeks), and all five children produce where questions with go,

going and gone before they produce incorrect goes questions. This suggests

that only after children have learned a series of separate frames associated

with different forms of Go from the input are they able to abstract the

construction Where’s X V to generate new utterances not learned directly

from the input. Moreover, although by this late stage in development the

children may have recognized a relation between forms of Go, it is clear that

they have yet to develop an understanding of how to mark tense correctly.

Input frequency and order of acquisition of Wh-questions. The input data

were examined to establish whether the frequency of different Wh-words

with different forms of Go predicted their order of acquisition in the

children’s speech. The mothers data from the first tapes were searched for

all Wh-questions and the frequency of use of each Wh-word with each form

of Go calculated. The mean frequencies of use of different Wh-words with

different forms of Go in the input (N¯ : forms of Go¬ Wh-words;% "±) were compared with their rank order of acquisition in thechildren’s speech. The rank order correlation was significant (ρ¯®±, df¯, p!±) suggesting that the frequency of use of different Wh-wordswith different forms of Go in the input is a good predictor of their order of

acquisition in the children’s speech.

The above analyses, based on a comparison of the use of individual lexical

forms in specific structures in the children’s speech and matched samples

from the input, show that even when sample size and knowledge of lexical

forms were controlled, the children showed less evidence of operating with

system-wide grammatical knowledge than their mothers. They used fewer

PP and infinitival complements with gone, and fewer Wh-words with

different forms of Go than were found in the input. Interestingly there was

some evidence of generalization between forms of Go (the use of Wh-

questions with goes and infinitivals with go), but crucially this occurred



ET AL.

relatively late in development, and was restricted to specific structures. The

late errors observed in the production of Wh-questions with goes suggest that

although children may have recognized a phonological similarity between

goes and other forms of Go, they have not yet developed a full understanding

of the grammatical role of this form.)

The present study aimed to provide a detailed picture of the acquisition of

the verb Go, and to examine the level of syntactic generality available to

children between the ages of ; to ;. A general analysis comparing the

children’s data for the whole year with the input data suggests that their use

of Go is largely adultlike. However, a lexical analysis taking a developmental

approach suggests that the children may not initially operate with a single,

adultlike representation of the verb Go in the lexicon.

With respect to the predictions derived from generativist and constructivist

accounts, the data suggest the following. First, the comparison of lexical

diversity in the children’s speech and the input showed that in specific

structures, with individual forms of Go, the children’s use of lexical forms

was more limited than that observed in the mothers’ data, even though there

was evidence that the children had acquired the relevant forms. These results

favour a constructivist interpretation of the data.

Second, the children did not produce specific structures, meanings, or

even lexical items (e.g. individual prepositional phrases) across forms of Go.

Instead, the order in which the structures and meanings of Go were acquired

reflected their relative frequencies in the input. These results can be

interpreted either as implicating a causal role for the input in determining the

pattern of acquisition, or as a reflection of sampling issues and the children’s

sensitivity to preferred forms of expression in English. Finally, the pattern of

errors that suggests a limited understanding of tense and agreement is not

[] One reviewer expressed concern that other errors might provide evidence for earlierintegration of forms of Go. However, an analysis of the children’s errors with the verb Go

shows that the vast majority of errors (%) consist of the use of an auxiliary form,predominantly contracted ‘s ’ with the form go (e.g. he’s go to look around ), with occasional

use with finite forms goes (e.g. that one’s goes there) and went (e.g. he’s went over his wheels).

These errors seem to reflect the children’s developing knowledge of subject–verb

agreement rather than their knowledge of the relation between different forms of Go,

especially given that these types of errors are also observed with other verbs in the

children’s speech. Other errors consist of the use of finite verb forms in Wh-questions

where finiteness should be marked by an auxiliary, and negation errors where finiteness

is incorrectly marked or omitted (don’t go}goes). Overall, the errors observed with Go seemto reflect the children’s lack of knowledge of the appropriate way to mark tense in different

argument structures, suggesting that their knowledge of the relation between forms of Go

is not fully integrated.



‘ ’

specifically predicted by generativist approaches, whereas constructivist

approaches predict that these types of errors will occur as children relate the

different forms of Go but sometimes enter them incorrectly into constructions

learned with other forms.

While the data are open to alternative interpretations, taken as a whole they

suggest that children’s early use of different forms of Go is lexically-specific,

and that the acquisition of the different structures and meanings associated

with Go occurs gradually and at different rates with different forms of Go.

Moreover, it appears that the children are not operating with an adultlike

representation of Go even by age ;, which suggests that children’s

linguistic systems remain organized around individual lexical items well

beyond the earliest stages of language acquisition. Although we favour a

constructivist interpretation of the data, in most of these analyses, we cannot

differentiate empirically between generativist and constructivist perspectives.

It is clear that the increasingly lexicalist nature of generativist accounts of

language acquisition, and their emphasis on competence rather than per-

formance may mean that ultimately, we are unable to differentiate empirically

between these approaches.

These data raise questions for both generativist and constructivist accounts

of language acquisition. If generativist theories are to explain these data, we

need explicit predictions as to how and when children build verb paradigms,

and to relate this process to the acquisition of tense and agreement marking.

Although many generative theorists would claim that building a unified verb

paradigm for Go will take time, many would also assume that children have

a knowledge of tense and agreement marking from very early in development.

These two apparently contradictory claims are not, to our knowledge,

currently integrated in any detail in generative theories of acquisition.

Second, generativists will need to develop new ways to test children’s

linguistic competence if they wish to assume that sampling issues or

performance factors reflecting preferred forms of expression can explain the

children’s lack of use of some structures and meanings with different forms

of Go. The traditional methods used to test children’s grammatical com-

petence, for example, examining the pattern of subject case marking with

finite and nonfinite verbs, are also open to alternative interpretations, because

the patterns of language use that generativists assume reflect an abstract

grammar are often predicted by constructivist accounts because they reflect

the patterns of language children hear.

However, to explain these data, constructivists also need to characterize

the process of acquisition in considerably more detail. For example, a lack of

evidence for the abstract grammatical representations, assumed by

generativist approaches to underpin children’s early speech, does not mean

that children only have lexically-specific constructions associated with

different forms of Go. Rather, there will be points in development where they



ET AL.

have neither very simple formulae, nor an adultlike grammar. The apparent

generalization of use of infinitival structures with going and go, and where

questions with goes, may relate to the abstraction of schemas (e.g. V to V,

Where’s X V?) from previously learned lexically-specific frames, perhaps on

the basis of distributional, phonological and}or semantic similarities. How-ever, we know very little about the precise circumstances under which these

abstractions are made, nor about the relative influences of input frequency

compared with semantic, distributional, or phonological similarity in sup-

porting abstraction. In this study, input frequency seemed to play a central

role in the acquisition process, irrespective of the specific relations between

forms of Go in terms of the structures in which they appear, their meanings,

or their phonological form. One possibility, however, is that the different

meanings associated with Go are organized around a prototype (e.g. move-

ment), and that the closer different meanings are to the prototype, the easier

they are for children to acquire and integrate into their existing linguistic

system. Similarly, some forms may be so different in meaning to the

prototype that they are stored separately, even in the adult grammar.

Another possibility is that forms that share phonological content (go}going}goes}gone) will be integrated earlier than forms that involve suppletion (went).However, we would need to differentiate the relative roles of frequency and

similarity in meaning or phonological form in much more detail to better

understand how abstraction in the system occurs.

A second problem for constructivist accounts lies in providing a more

detailed characterization of adult linguistic representations. These data are

compatible with usage-based models of adult language (Langacker, ;

Croft, , in press a), in which linguistic knowledge is organized around

constructions rather than the abstract grammatical concepts that form the

basis of generativist models. Usage-based models assume that adults store

linguistic information at a number of levels, from the lexically-specific to the

more abstract, and frequency of use determines whether utterances are stored

as linguistic wholes or constructed from more abstract units (Bybee, ).

Therefore, in the adult system, the frequency of use of forms of Go with

different structures and meanings will influence whether they are stored as

lexically-specific constructions, or generated from more abstract schemas.

Thus, adults as well as children are thought to use lexically-based schemas

to produce many of their utterances. However, because the extraction of

abstract schemas that provide links between lexically-specific schemas is

thought to be based on accumulating linguistic experience, and to take place

in development, children will initially have only highly specified con-

structions based around high frequency lexical items. The difficulty is that

without a detailed characterization of how different levels of information are

represented in the adult system, we know very little about how much



‘ ’

development is needed in children’s systems to reach an adultlike knowledge

of grammar.

Aside from the limitations outlined above in providing a satisfactory

account of these data, we suggest that researchers should be cautious in

crediting children with an abstract knowledge of tense marking. Although

only one verb was examined, Go is the most frequent verb, and goes the most

frequent rd person singular verb (excluding the copula) produced by of

the children (nd most frequent for children), accounting for a mean of

±% (range ±–±%) of all rd person singular verb use, and isarguably the verb about which they know the most. Yet these data suggest

that children’s knowledge of the relations between different forms of Go may

be less abstract than is often assumed (based on an analysis of over uses).

It is therefore unclear what conclusions regarding children’s grammatical

knowledge should be drawn from the sporadic use of other rd person

singular forms many of which are produced just once in the speech of

individual children. If go and goes are initially learned as unrelated lexically-

based constructions, this would count as evidence against an ‘Optional

Infinitive’ interpretation of early verb use (Wexler, ) because it is

unclear in what sense the use of tense can be said to be ‘optional ’.

In conclusion, the detailed picture of early verb use documented in this

study is broadly consistent with a constructivist approach to early language

acquisition. The data suggest that children build up adultlike grammatical

knowledge by beginning with specific lexical frames and gradually moving

towards more abstract schemas. However, much more work is needed to

establish how children form schemas which operate across larger groups of

lexical items. In addition, this study has demonstrated the difficulties

associated with attempting to derive empirically testable predictions that

differentiate between generativist and constructivist approaches to early

language use. Where it is impossible to differentiate empirically between the

two positions, it is a matter of theoretical preference which theory is adopted.

Further analyses incorporating a wider range of verbs are required to enable

us to better understand the nature of children’s early understanding of tense

and agreement and the interaction between their knowledge of these systems

and their use of syntax. It appears, however, that by examining the input

with respect to individual lexical items rather than in terms of abstract

grammatical categories, it is possible to uncover relationships between

children’s language use and the language they hear which raises interesting

questions as to the precise role of the input in acquisition (Lieven, ;

Pine, ; Rowland & Pine, ; Theakston et al., ).



ET AL.

REFERENCES

Akhtar, N. & Tomasello, M. (). Young children’s productivity with word order and verbmorphology. Developmental Psychology , –.

Bloom, L., Miller, P. & Hood, L. (). Variation and reduction as aspects of competencein language development. In A. D. Pick (ed.), Minnesota Symposia on Child Psychology,

Vol. . Minneapolis : University of Minnesota Press.Bybee, J. (). Regular morphology and the lexicon. Language and Cognitive Processes ,

–.Bybee, J. (). The emergent lexicon. In M. C. Gruber, D. Higgins, K. S. Olson & T.

Wysocki (eds), Papers from the th annual meeting of the Chicago Linguistics Society, Part: the panels. Chicago: Chicago Linguistics Society.

Bybee, J., Perkins, R. & Pagliuca, W. (). The evolution of grammar. Chicago: TheUniversity of Chicago Press.

Bybee, J. & Scheibman, J. (). The effect of usage on degrees of constituency: thereduction of don’t in English. Linguistics , –.

Clark, E. (). Discovering what words can do. In D. Farkas, W. Jacobsen & K. Todrys(eds), Papers from the parasession on the lexicon. Chicago: Chicago Linguistic Society.

Croft, W. (). Linguistic evidence and mental representations. Cognitive Linguistics ,–.

Croft, W. (). Some contributions of typology to cognitive linguistics (and vice versa). InT. Janssen & G. Redeker (eds), Cognitive linguistics: foundations, scope and methodology.

Berlin: Mouton de Gruyter.

Croft, W. (in press a). Radical construction grammer: syntactic theory in typological perspective.

Oxford: Oxford University Press.

Croft, W. (in press b). Lexical rules vs. constructions: a false dichotomy. In H. Cuyckens, T.

Berg, R. Dirven & K. Panther (eds), Motivation in language: studies in honous of Guenter

Radden. Amsterdam: John Benjamins.

Goldberg, A. E. (). Constructions: a construction grammar approach to argument structure.Chicago: The University of Chicago Press.

Langacker, R. W. (). Foundations of Cognitive Grammar, Vol. : Theoretical Pre-requisites. Stanford, California : Stanford University Press.

Langacker, R. W. (). Foundations of Cognitive Grammar, Vol. : Descriptive Application.Stanford, California : Stanford University Press.

Levin, B. & Hovav, M. Rappaport (). Unaccusativity: at the syntax-lexical semanticsinterface. Cambridge, MA: MIT Press.

Lieven, E. V. M. (). Crosslinguistic and crosscultural aspects of language addressed tochildren. In C. Gallaway & B. Richards (eds), Input and interaction in language acquisition.

Cambridge: CUP.

MacWhinney, B. (). The CHILDES Project: tools for analyzing talk. Hillsdale, NJ:Erlbaum.

MacWhinney, B. & Snow, C. (). The Child Language Data Exchange System: an update.Journal of Child Language , –.

Marcus, G., Pinker, S., Ullman, M., Hollander, M., Rosen, T. & Xu., F. ().Overregularisations in language acquisition. Monographs of the society for research in child

development (Serial no. , Vol. ).Naigles, L. R. & Hoff-Ginsberg, E. (). Why are some verbs learned before other verbs?

Effects of input frequency and structure on children’s early verb use. Journal of Child

Language , –.Pine, J. M. (). The language of primary caregivers. In C. Gallaway & B. Richards (eds),

Input and interaction in language acquisition. Cambridge: CUP.

Pine, J. M., Lieven, E. V. M. & Rowland, C. F. (). Comparing different models of thedevelopment of the English verb category. Linguistics , –.

Pinker, S. (). Language learnability and language development. Cambridge, MA: HarvardUniversity Press.



‘ ’

Pinker, S. (). Learnability and cognition: the acquisition of verb-argument structure.Cambridge, MA: Harvard University Press.

Pinker, S. (). The language instinct. London: Penguin.Radford, A. (). Syntactic theory and the acquisition of English syntax. Oxford: Blackwell.Rowland, C. F. & Pine, J. M. (). Subject–auxiliary inversion errors and wh-question

acquisition: ‘what children do know?’. Journal of Child Language , –.Theakston, A. L., Lieven, E. V. M., Pine, J. M. & Rowland, C. F. (). The role of

performance limitations in the acquisition of verb-argument structure: an alternative

account. Journal of Child Language , –.Tomasello, M. (). First verbs: a case study of early grammatical development. Cambridge:

CUP.

Wexler, K. (). Optional Infinitives, head movement and the economy of derivations. InD. Lightfoot & N. Hornstein (eds), Verb movement. Cambridge: CUP.

Wexler, K. (). Very early parameter setting and the unique checking constraint : a newexplanation of the optional infinitive stage. Lingua , –.

journal of child language €¦ · downloaded: 19 feb 2016 ip address: 194.94.96.194 theakston et...

Documents